Blockchain and Machine Learning in Health

Care and Management

Sankarr Jain

Department of Electrical and

Electronics Engineering

Ramaiah Institute of Technology

Bangalore,India

sanskar329@gmail.com

Kushagra Awasthi

Department of Electronics and

Communication Engineering

Ramaiah Institute of Technology

Bangalore,India

kushagra.awasthi33@gmail.com

Aditya Anand

Department of Electronics and

Instrumentation Engineering

Ramaiah Institute of Technology

Bangalore,India

aditya.anand556@gmail.com

Sarvesh Gujrati

Department of Mechanical

Engineering

Ramaiah Institute of Technology

Bangalore,India

srvshgujarati3@gmail.com

Aman Gupta

Department of Information

Science and Engineering

Ramaiah Institute of Technology

Bangalore,India

gupta.aman1602@gmail.com

Janamejaya Channegowda

Department of Electronics and

Electrical Engineering

Ramaiah Institute of Technology

Bangalore,India

jc@msrit.edu

Abstract— Today we have enormous amount of data available in

every sector, with the advent of technology available, it is possible

to provide solutions to many problems. In this paper we are going

to provide solutions to the problems related to healthcare data

management using Machine Learning and Blockchain. Extracting

only the relevant information from the data is possible with the

use of Machine Learning. This is done using trained algorithms.

Once this data is stored, the next problem is Data sharing and its

reliability. This is where Blockchain comes into picture. The

consensus in Blockchain technology makes sure that data is

legitimate and transactions are secure. Blockchain technology can

potentially change health care management for the better by

placing patient at the epicentre of the healthcare system and

increasing the privacy and interoperability of health data. This

paper focuses primarily on solving healthcare data

management problems by using Blockchain technology and

including some indispensable features using Machine Learning.

Keywords— Bag of words, blockchain, Electronic Health

Records (EHR), Machine Learning, Social Security Numbers

(SSNs).

I. INTRODUCTION

In Blockchain like a distributed ledger, singular

transactions are encrypted into blocks by the applicable

encryption, added to the ledger and never deleted. The

information in Blockchain is verified fundamentally by a

linked list of encoded exchanges that utilizes a hash. The hash

function generates a hash by encrypting the information fed

in Blockchain. It shapes the foundation of a decentralized

medicinal service stage shared by the patients and suppliers,

acting as an interface to the patient's record. [1]

Blockchain is a cryptographically secured, immutable,

write once, read anywhere type data structure. It consists of

blocks and theses blocks are linked together using an

unmodifiable key referencing mechanism.

Fig.1 Blockchain Networking model

The Blockchain data structure consists of the following

components:

 The Blockchain network has secured list of blocks

which contains the useful information.

 A peer-to-peer network which contains identical

examples of the Blockchain data structure

 A consensus mechanism which secures the

harmonized growth of Blockchain.

 A security mechanism that ensures that the data sto

red in the Blockchain network is immutable.

II. MEDICAL DATA BREACH

Reportedly, In April 2019, healthcare data breaches

reached a record high and for the past few years, the

healthcare field has had the second highest number of

breaches compared to other sectors. In total, almost 10

million records were exposed in US healthcare sector in 2018

alone. The frequency of medical data breaches has

been highly concerning. In particular, armed with someone’s

medical information, thieves can easily commit medical

identity theft to get drug prescriptions, or make false

insurance claims under the victim’s name. Medical data

mostly comes with personal and private information which

includes Social Security Numbers (SSNs), as well as

financial information. These pieces of data can be easily sold

on the black market or used for malicious crimes, including

credit card fraud and identity thefts. Criminal with such

sensitive information can also file a fraudulent tax returns

using the Social Security Number. [3]

The healthcare data breach statistics that was issued by

the Department of Health and Human Services’ Office for

Civil Rights, U.S., clearly indicates the threat of data breach

in the present scenario. This data was compiled for the decade

starting from 2009 to 2018. The healthcare data breach

statistics as shown in the above figure shows that the year

2018 has seen the highest amount of data breach. Also, in the

last decade there has been about 2,546 healthcare data

breaches involving more than 500 records, which was the

cause of exposure of not less than 189,945,874 healthcare

records equating to more than 59% of the population of the

U.S. [3]. The healthcare data breach statistics as shown in the

above figure shows that the year 2018 has seen the highest

amount of data breach

Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.

Fig.2 Number data breaches over the years [4]

Also, in the last decade there has been about 2,546

healthcare data breaches involving more than 500 records,

which was the cause of exposure of not less than 189,945,874

healthcare records equating to more than 59% of the

population of the U.S. [3]. The next statistics shown in fig 3

indicates that Hacking and IT incidents have caused majority

of these data breaches.

Fig.3 Number of hacking incidents in the past decade [4]

In many cases, leakage of data can prove to be life

threatening and may cause serious safety problems to many

innocents. In India specifically, there have been many

incidents of medical report leakage. Some of them are listed

here:

 In April 2018, Andhra Pradesh government

websites were leaking Aadhaar number and

sensitive information like reproductive history

from pregnancy to delivery and details

about abortions and so on, The Times of India

reported.

 In 2018, an unsecured Andhra Pradesh government

website exposed the names and contact numbers of

every person who bought medicines which also

includes the customers who bought medicine for

erectile dysfunction from government-run stores.

The dashboard on the Anna Sanjivini website

unintentionally grated access to details which

includes the names and phone numbers of every

individual who bought medicines from every store.

[5]

When sensitive information is leaked online, it can lead

to malevolent consequences, also having the potential of

increasing the crime rate significantly.

III.DECISION MAKING ERROR

Another problem is the transfer of data. clinicians heavily

rely upon suitable clean data to make decisions regarding

possible treatments that can resolve the health issue in the

least possible time. It is important to have a secure means to

transfer patient’s historical data to hasten the treatment

process.

In literature, studies have been carried out to estimate the

number of deaths related to medical error. The American

Johns Hopkins Hospital by Makary et al have concluded

that deaths due to Medical Error have significantly increased

over the years in USA. These statistics are shown in Fig.4

Fig.4 Leading cause of deaths in USA [6]

IV. STORING MEDICAL INFORMATION

Though there has been some advancement in the way

medicals reports are stored, their transactions and security are

still not up to the mark. Health data breaches are frequent and

with the subsequent increase of hackers all around the globe,

hacking of these data has become predominant,

overshadowing the other breaching methods.

Meditab, a software company, claims itself as one of the

foremost electronic medical records software creators for

medical institutions. The company includes electronic faxes

and this is method is still used significantly to share patient

data to other data seekers. But this method of sharing data has

proved to be insecure and unreliable making it very less

preferable.

Spider Silk, a Dubai-based cybersecurity firm, had a fax

server, which was running an Elastic search database which

included not less than 6 million health care records .The

server did not even had passwords for security, which

indirectly granted access to everyone and thus anyone could

read the transmitted faxes in real-time .The fax that were sent

consisted of many recognizable information regarding

patients such as their medical history, treatment undertaken

in the past, their Social Security Numbers and other records.

[7]

V.ELECTRONIC HEALTH RECORD [EHR] AND ITS

LIMITATIONS

Keeping in mind the issues with Health record

management that was used earlier, Electronic Health

Record (EHRs) were introduced. EHR’s are electronic

version of a patient's medical history, which is maintained by

the provider over time, and includes all important

organisational clinical data which concerns the patient.

Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.

EHRs consist of several information management

strategies for various purposes. There have been numerous

data breaches cases reported which thereby confirms that the

security of these reports has to be increased further.

There have been many cases where EHRs have failed in

the area of security. Listed below are some incidents which

proves this point. [8]

 Patient records were leaked in University of Michigan

Medical Center patient records were disclosed on the internet

the password.

 In Florida, a public health employee was able to access

thousands of names of HIV infected patients and he sent it to

many local news stations

 The University of Washington Medical network facility

was accessed by unauthorized personnel and patient data was

compromised

 Medical centres in New York and Holland and their

weakness in protecting patient data was exposed

 A group of researchers from University of Minnesota

inadvertently released names of kidney donors to the public.

Exposing a patient’s private information to the general

public can have disastrous effects on the reputation and

mental health of the patient. This sort of information leak can

also lead to financial losses to the individual involved.

The cases mentioned clearly indicates the unintended

problem that comes up with the usage of EHRs. The Health

care model proposed in this paper deals with many of these

problems and provides meaningful solutions. [9]

In our proposed model we provide solution to almost

all of the problems mentioned above.

VI. OUR PROPOSED MODEL

In our proposed model we have used both Blockchain

technology and Machine Learning Algorithms to provide a

better solution in terms of security. By using Machine

Learning, we provide additional features which can be the

base of ideas for further implementations on this subject.

Machine Learning is based on the concept of

centralization of data, while Blockchain technology uses

decentralization of data to provide high security. In this paper

we have tried to project our model which showcases how we

can use both these for this particular application.

A. Implementation of Blockchain

In this model, dual Blockchain structure is used, the first

part grants access to health data and is built using the

Hyperledger Fabric. The second part of the structure works

on Ethereum and performs all application and services.

Medical information is very sensitive and personal so a

closed Blockchain such as Hyperledger Fabric helps in

retaining necessary privacy required.

Majorly blockchains are classified as public Blockchains

and permissioned Blockchains.

This can be explained by considering the example of a

user wants who to sell a book to person with some rebate and

does not intend to tell about this to general public, the seller

then can employ permissioned Blockchain to hide the

information about the offer from the public. This model uses

a double encryption mechanism on a permission-based

Blockchain. The security that is provided by this model

which uses Blockchain is beyond and far more advanced than

any other centralized security system being used.

Furthermore, the patient’s data is made inaccessible and

unalterable. The Blockchain acts as a pointer, and provides

the direction to the location of the stored data, meaning that

anyone attempting to access patient data will be denied.

The health data is secured between the patient and the

authorized doctor. When the authorised doctor adds

additional information to the patients record history, the

system will automatically update it. Only those clinicians

who have authorized access can view the updates. None of

the doctors are given permanent authorization, the access for

the doctor ends when the patient wants so that the doctor can

no longer update the record or access it.

This is vital in scenarios where there is a need to change

the doctor in charge, therefore, with the help of Blockchain

the information transfer will be easy and secure. Issues

associated with the transfer of information by medical

institute employees is completely eliminated and there will

be no more data leaks probable in this transaction process and

also overseas transactions of information can be cost

effective as compared to the conventional techniques to do

the same.

In emergency situations when the patient is unconscious

and unable to provide any sort of input on his health, it would

be vital to have access to the patients’ health records. This

information while performing lifesaving surgeries as history

of past medications and illnesses are crucial before

performing any sort of major surgery.

B. Implementation of Machine Learning

There are two steps in building a new Machine learning

model. The first step is to take in the dataset and adjust the

model weights to increase the accuracy of the model. The

second step is testing the Machine learning model on

independent or new datasets for the accuracy of the model and

thus validate the model and prevent overfitting of the model.

An over fitted model is very good at a given dataset but is bad

at hypothesizing for the given problem. The procedure of

Machine learning in our proposed model has been depicted in

Fig.5.

Once a Machine learning model has been trained using

Supervised Learning it can be used to do various tasks such

as prediction, classification on the untrained dataset. In the

proposed model, we use the “Bag of Words” algorithm

which will extract only the required dataset and ignore the

various other things like the Name, Age, Address and other

personal details of the patient to maintain the privacy.

Supervised Learning builds a mathematical model of the

given dataset that contains both input and output data.

Supervised Learning can be used for classification and

regression.

Steps for Supervised Learning:

 First, the category of training set data is selected

 The set of input data and its complimentary outputs are

collected.

 The accuracy of the trained function depends on how the

input dataset is represented. Typically, the input dataset is

Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.

converted into a feature vector, which contains a number

of features that describe the dataset. The number of

features should not be too large, because of the curse of

dimensionality; but should contain enough information to

accurately predict the output.

 We need to determine the corresponding algorithms

according to the given dataset.

 Run the algorithm being trained on the collected training

set.

 The output of the test dataset determines the accuracy of

the algorithm.

There are four major issues to be considered in Supervised

learning such as Bias- Variance, Trade Off, Function

complexity and amount of training data, Dimensionality of

the input space.

Fig.5 Supervised learning flowchart [10]

To get rid of some of the above issues we use the “Bag

of Words” algorithm which is a method of extracting

required features from text for use in modelling. The trained

data contains terms in this algorithm and the correct features

are selected from these words.

Upon defining the features, the feature is given a specific va

lue, this feature representation is mapped to the training and

test dataset. For the reflection of the BOW the two most

common features value representation are:

1) Binary feature value:

The value of this representation feature can be either 0 or 1

where 1 indicates that the feature is in the dataset and 0 indi

cates that the feature is not in the dataset.

2) Frequency Feature Value:

The value of a particular feature in this representation

is determined by the number of times it appears in the datase

t and if it did not appear, it is 0.

In this paper, we are using frequency feature value

representation. A feature is considered important in this

representation if it appear more than once in a sentence.

Identifying and deleting phrases, that do not include illness

or treatment information makes the data set more effective.

The proposed algorithm extract from the HTML page only

disease related information containing medical information

and neglects other irrelevant materials such as forms, users,

commentaries, navigation, menus, feedback and advertising.

The benefit of using this method is that by using a weighted

Bag of Words algorithm it improves precision with an

accuracy of 79% to 82%. [11]

This model will take the dataset from reliable sources like

Medline which provides the Machine learning model with the

required information, which filters it into specific categories.

Using health data of patients with diseases that are very rare

is also of great importance to the medical scientists and

researchers. Consider, for example, a case where there is an

outbreak of a new disease in a particular area or country, about

which there isn’t much information available. Machine

learning plays a vital role in such circumstances where the

information produced by the health reports of the patient is

fed in the form of data to the cloud without allowing access to

the personal and private information of the patient which in

turn will help the researchers and the doctors with some

reliable and accurate information with better analysis of the

symptoms and effects of the disease using the Machine

learning technology.

When the health data of a lot of patients is fed to the

machine, it provides enough statistical information about a

particular disease or disorder. Moreover, categorization of

diseases will make it more convenient for a patient to access

and understand their respective health records and will also

facilitate researchers in searching for the information.

It is important for the patients to have control over their

health record and also benefit from the data that they share.

Therefore, it is possible to find users willing to have their

health data used for research. These patients will be given

clear information about the exact purpose for which the

information will be used and also what kind of data will be

required. As the data is anonymized, there will be no issues

of privacy. In return, participants can be compensated.

C. Blockchain and Machine Learning

Blockchain used along with Machine learning opens up a

host of opportunities for securing health data. It is relatively

strenuous task to combine both of them. There has not been

significant work done in this area.

In this paper, we have used Supervised learning to train

the Machine learning algorithm on the datasets obtained

through various sources like MedLine, and to reduce the

dimensionality of the data we will use the “bag of words”

algorithm which will take only the necessary parts of the

dataset to train the Machine learning model.

The Patient healthcare data is secured using Blockchain

network through which transactions between patient and the

authorized doctor is made. The functionality of Blockchain

ends here.

A new set of patient healthcare data is then fed to the

trained model which first filters the data and discards all the

personal details and information, then this data is further

categorized by diseases. The role of the Machine Learning

terminates here.

Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.

This dataset from a hospital can be transferred to other

hospitals by using e-mails or by uploading the data set from

each hospital on a cloud-based storage system which can be

accessed by hospitals and researchers.

Fig.6 Implementation of the proposed model

This sharing of data is very important because the data

for each hospital would be completely different. The biggest

advantage of using Machine Learning algorithms is to make

the health care data accessible to researchers and doctors for

study purpose.

Thus, using this model can provide dual advantages with

Blockchain providing the necessary security

of the healthcare record by allowing secure transactions

between the doctor and patient, while the filtered data can be

used by researchers and doctors because this data need not be

secured as it is deprived of all personal information. This data

is entirely built to be used for reference purpose.

The above application is performed by Machine Learning

algorithm as discussed above.

VII. CONCLUSION

Blockchain Technology has been evolving with time,

banking and financial sectors are already using Blockchain

keeping in mind the unparalleled advantages it offers, with

the significant increase in health data breach through

hacking, and application of Blockchain for security

becomes important and imperative. It will not be wrong to

say that Blockchain based Health care model is the future in

healthcare sector and has the potential to change the way

health care records are managed and secured.

With the emergence of 5-G networks and faster than ever

data transfer facilities it will encourage advancement of

Machine Learning, Blockchain and other data based

techniques in various sectors including Healthcare. As this

new technology ecosystem emerges, Blockchain promises

significant improvements in managing patient health records.

Continuous efforts are being made to increase the

accuracy of wearable health tracking devices and if these

data could provide more accurate and reliable results there

will be brighter chances of integrating these devices with the

health records to provide more information and also share

some of these medical data securely with authorized doctor

without actually visiting.

The ideas based on implementing Blockchain and

Machine Learning is not much explored. Our paper proposes

a unique healthcare model which is still in its infant stage but

can surely form base to many more health care models to

come.

There is further scope of improvising this idea by

implementing Artificial Intelligence (AI), Internet of things

(IoT) and much more available technology to develop a more

comprehensive health care model in future.

REFERENCE

[1] Chen, Y.; Ding, S.; Xu, Z.; Zheng, H.; Yang, S. "Blockchain-Based

Medical Records Secure Storage and Medical Service Framework",

Journal of Medical Systems, vol.43, no. 5, 2018.

[2] G.Magyar,Blockchain: Solving the privacy and research availability

tradeoff for EHR data: A new disruptive technology in health

data management, Budapest, Hungary,24-25 Nov. 2017.

[3] Aimee O’Driscoll (2019, July), “The biggest medical data breaches in

history”,

a. [Online].Available: https://www.comparitech.com/blog/vpn-

privacy/biggest-medical-data-breaches/

[4] HIPAA JOURNAL (2018,March) , “Healthcare Data Breach

Statistics.”,

a. [Online]. Available:

b. “https://www.hipaajournal.com/healthcare-data-breach-

statistics/

[5] Zaheer Merchant (2019, April) , “Health department of northern

state exposed data of 12.5 million pregnant women.”,

a. [Online].Available: https://www.medianama.com/2019/04/223-

health-department-indian-state-pregnant-women-data-leak/

[6] "Medicalchain Whitepaper - 2.1"

a. [Online].Available: https://medicalchain.com/en/whitepaper/

[7] Zack Whittaker (2019, March), “A huge trove of medical records and

prescriptions found exposed.”,

a. [Online]. Available:

b. https://techcrunch.com/2019/03/17/medical-health-data-leak/

[8] Centres for Medicare & Medicaid Services,” Electronic Health

Records”,

a. [Online]. Available:

b. https://www.cms.gov › Medicare › E-Health › EHealthRecords

[9] K.T. Win. A review of security of electronic health records.

Electronic Health Records: security, safety and archiving, 34, 2005

[10] .S. B. Kotsiantis,Supervised Machine Learning: A Review of

Classification Techniques, Department of Computer Science and

Technology University of Peloponnese, Greece.

[11] P. Shinde,S.Madhav,Health Analysis System using Machine Learning

Proceedings of the International Conference on Mainstreaming Block Chain Implementation (ICOMBI) 2020

Authorized licensed use limited to: STATE UNIV NY BINGHAMTON. Downloaded on August 31,2023 at 22:22:32 UTC from IEEE Xplore. Restrictions apply.